← Back to PEARL AI Voice

Privacy and licensing

Data processing

The chosen browser speech engine processes entered text on the visitor's device after downloading models. The PHP landing page does not transmit text to a speech API. Remote model repositories may receive normal download requests, including IP addresses, subject to their privacy practices.

Third-party models and code

The installed engine must carry applicable attribution and model licenses. Review the upstream Svāra WebGPU code repository, the Svāra ONNX weights, SNAC codec, Transformers.js, and transitive dependencies before commercial release. This distribution does not include third-party code or model weights.

Outputs and acceptable use

Users should have the right to process entered text. No unauthorized impersonation, voice cloning or unlawful distribution.

Platform availability

Large model downloads and GPU resources are required. No guarantee of performance on all devices.